HAL Id: hal-00190694https://telearn.archives-ouvertes.fr/hal-00190694

Submitted on 23 Nov 2007

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

KMsim: A Meta-Modelling Approach and Environmentfor Creating Process-Oriented Knowledge Management

SimulationsAnjo Anjewierden, Irina Shostak, Robert de Hoog

To cite this version:Anjo Anjewierden, Irina Shostak, Robert de Hoog. KMsim: A Meta-Modelling Approach and Envi-ronment for Creating Process-Oriented Knowledge Management Simulations. 13th European Confer-ence on Knowledge Acquisition, Management and Modelling (EKAW), 2002, Siguenza, Spain. 15 p.�hal-00190694�

KMsim: A Meta-Modelling Approach and Environmentfor Creating Process-Oriented Knowledge Management

Simulations

Anjo Anjewierden1, Irina Shostak2, and Robert de Hoog1,2

1 Social Science Informatics, University of Amsterdam,Roetersstraat 15, 1018 WB Amsterdam, The Netherlands,anjo@swi.psy.uva.nl

2 Faculty of Educational Science and Technology, University of Twente,PO Box 217, 7500 AE Enschede, The Netherlands,{shostak,hoog }@edte.utwente.nl

Abstract. This paper presents a new approach to modelling process-orientedknowledge management (KM) and describes a simulation environment (calledKM SIM) that embodies the approach. Since the beginning of modelling researchershave been looking for better and novel ways to model systems and to use appro-priate software to create simulations. The application of the approach and KM-SIM make it possible to create realistic business models (BMs) and simulate theconsequences of KM interventions and events. The validity of the approach andtools is being evaluated in the game KM Quest.

1 Introduction

With the ever growing interest for knowledge management, it is unavoidable that thedemand for a more formal approach increases in parallel. After the first flush of ideas,whose main function it was to create awareness, more precise and hands-on methodsare called for (see for example [9]). This holds in particular for models that show howknowledge and knowledge processes can influence organisational effectiveness (see [5],[4] and [8]). This “show how” becomes even more valuable when these influences canbe simulated in a business model (BM), as this is the only way one can capture andunderstand the dynamics of knowledge. The need for modelling and simulating knowl-edge management relevant business models raises the question whether additional toolsare required beyond the standard simulation environments already available.

This paper describes KMSIM, a set of tools which have been specifically designedto support creating and simulating knowledge management relevant business models. Itis argued that the need for these tools can be derived from the nature of knowledge man-agement as a discipline, the peculiar properties of knowledge relevant business modelsand the intended users of the tools. The tools were developed in the context of theKITS project. The goal of this project is to develop a comprehensive game-based col-laborative learning environment for knowledge management called KM Quest [1]. Anessential part of this environment is a knowledge management relevant business modelthat simulates the behaviour of a (fictitious) company. This model has been developedand partly validated with the tools described in this paper.

The paper consists of four sections. In Sect. 2 the factors driving the need for aspecific and new set of tools are discussed. Sect. 3 describes the architecture of thesimulation environment based on the requirements. The last two sections describe thefunctionality provided by the tools from the point of view of creating business models(Sect. 4) and simulating and validating these models (Sect. 5).

AcknowledgementsWe would like to thank the three anonymous reviewers whosecomments we have tried to take into account. Work partially supported by the EuropeanCommunity under the Information Society Technology (IST) RTD programme, con-tract IST-1999-13078 (KITS). The authors are solely responsible for the content of thisarticle. It does not represent the opinion of the European Community, and the EuropeanCommunity is not responsible for any use that might be made of data appearing herein.Partners in the KITS project are University of Twente (NL), University of Amsterdam(NL), CIBIT (NL), ECLO (UK), Tecnopolis (I) and EADS (F).

2 Factors Driving the Design of the Tools

2.1 Knowledge Management has an “Object”

Knowledge management, as a branch of general management disciplines, has an “ob-ject” that is different from the “objects” that are the focus of other sub-disciplines. Thereare many simulation environments that allow one to model various kinds of businessprocesses, including manufacturing, public systems, and service systems simulations(for example PowersimR©). Most environments, however, do not provide for treatingknowledge as asimulated entity. Rather, in these environments the simulated entitiesare theimplicit result of applying knowledge in a specific domain. Knowledge is notconsidered an object on which different actions can be applied, for example to model“stocks” and “flows” of knowledge.

To illustrate this idea, consider a manufacturing simulation which allows one to getanswers to questions like “How can work-in-process inventory and cycle time be re-duced while increasing throughput?” or “When should the next piece of equipment bepurchased and how many people are needed to work with this equipment?” In this simu-lation knowledge about manufacturing processes isappliedwhile simulating inventory,amount of labour, time arenot taken as a simulated entityin contrast to knowledge man-agement simulations. “Stocks” and “flows” of domain-specific knowledge compose thearea of interest of KM simulations. For KM simulations it is important to quantify, mea-sure and model “manufacturing knowledge” as a simulated entity, which can be done byintroducing variables such as level of competence in manufacturing and speed of knowl-edge gain in manufacturing. As an example of one of the first knowledge managementsimulations we can mentionTango![2]. Apparently the business model underlying thissimulation does not handle knowledge as a separate entity, but operates directly throughemployees on key performance indicators of a company. So, the nature of the object ofknowledge management in terms of stocks and flows requires a set of tools that al-lows the modeller of the knowledge management relevant business to map the “paper”representation of this model with a minimum of effort on a simulation engine.

2.2 Nature of the Knowledge Management Relevant Business Model

The model that is introduced in this paper was developed for KM training and is appliedin the game KM Quest. To support this function the business model should satisfy thefollowing basic principles:

– A business units’ output depends on thelevel of knowledgeand thelevel of knowl-edge usage(or utilisation). The output of work would be more valuable if people inthe company possess better and use/apply (more recent, novel, advanced) knowl-edge. However this result could be counteracted by non-effective organisation ofwork processes, and vice versa. The ideal situation consists of effective organisa-tion of work processes and highly skilled, highly knowledgeable employees.

– Certain changes outside or inside a company should influence its knowledge house-hold – individual and organisational knowledge.

– An important assumption is that knowledge can be seen as a quantifiable object andcan be measured in relative terms.

– There is a natural depreciation of knowledge due to volatility, instability, and ageingof knowledge. If in a company nobody takes care of renewal, gaining, and retentionof knowledge, the company will in the long run not be able to compete with marketconditions.

These considerations play a fundamental role in our modelling approach. Thus, inthe model knowledge “stocks” are introduced as the level of competence(s) and knowl-edge “flows” are introduced as the efficiency of processes involving knowledge, suchas knowledge gaining, development, utilisation, transfer, and retention.

Simply stated, any event that happens outside or inside the company and any inter-ventions taken inside the company can have an influence on the knowledge processes– knowledge flows. These also influence the “state” of knowledge in the organisation –knowledge stocks, which influence business processes and determine their quality. Fi-nally, the business processes contribute and generate the values of the key organisationaleffectiveness variables like profit and market share.

Based on what has been said above, our modelling approach resulted in a four-levelmodel consisting of:

Organisational effectiveness variablesThese variables reflect the competitive char-acteristics of the company and are represented by variables like market share, profit,level of sales and so on.

Business processes related variablesThese reflect the quality of internal processesand “how well” work is done within the company. Examples are: average time ofbringing a new product to the market and production level.

Knowledge related variables These variables reflect the relevant knowledge domains(e.g. marketing, research, production) and represent the level of competence in eachdomain.

Knowledge processes related variablesReflect the properties of processes involvingknowledge in the organisation (e.g., speed of knowledge gaining, effectiveness ofknowledge transfer).

Organisational effectiveness variables

Business process related variables

Knowledge related variables

Knowledge process related variables

Internalevents

KMinterventions

Externalevents

Fig. 1.Conceptual structure of the business model

The general structure of this knowledge management relevant business model isshown in Fig. 1.

Events and interventions are important components of the model. In our view, eventsare any changes outside or inside a company that happen independently from manage-ment of a company. Interventions are actions taken by management in order to preventor to react to events and are aimed at improving theknowledge householdof a company.These interventions becomeknowledge managementinterventions and differ from man-agerial interventions by its operational object(s).

However, knowledge is something that is difficult to measure in absolute terms. So,it is impossible to perform actions that (immediately) increase the “amount” of knowl-edge, i.e. quantitative characteristics of knowledge. On the other hand, it is relevant andpossible to change the quantity of other objects such as raw materials, time, amount oflabour, investments, which are the subjects of other managerial interventions. One canargue that managerial decisions concern not only these tangible, physical objects, butalso include decisions about strategic development, market policy and strategies, part-nership policy, and so on. Those decisions are qualitative and based on and impactingknowledge that is needed for a company to improve its value. Simpler examples includethe decisions to conduct training programmes or ICT implementation. Those decisionslead to qualitative changes in the organisation and, in many cases, cannot be measureddirectly and more importantly they affect theknowledge household.

As a consequence we should treat interventions not as a quantitative, but as qualita-tive entities and which require a very specific implementation in the simulation.

Summarizing what was discussed above and referring to the classification of models[6], and the types of interactions that can occur between discretely changing and con-tinuously changing state variables [7], both discrete and continuous components mustbe present in our simulation model, in particular the ones listed below:

– State variables change continuously with respect to time. Knowledge related vari-ables exhibit decay behaviour and consequently influence the state of other typesof variables;

– Discrete events (in our terminology - events and KM interventions) cause discretechanges in the value of continuous state variables;

– Continuous state variables achieving threshold values may cause a discrete eventto occur. Threshold values of knowledge related variables could be conditioned toenable occurrence of several events. This feature of the model is relevant for thegame and probably not applicable in reality, since events are unpredictable in manycases. Despite this fact, events still can be generated to consider several scenarios.

In addition we assume that the state of the business is never monitored on a perma-nent basis as is done both on aircraft and in many industrial processes. Usually somekind of reporting takes place at fixed points in time (monthly, quarterly, yearly). Thisshould also be reflected in the model: it should be able to provide reports about relevantvariables at pre-determined time intervals.

2.3 Practical Requirements for Tool Support

Apart from factors derived from the topic (knowledge management) and the businessmodel, also factors reflecting the intended users of a tool are important. In general,the quality of a tool depends to a large extent on how the vocabulary of the user ismade available in terms of tool functionality. The operationalisation of the functionalityshould be hidden from the user as much as possible.

In the KM Quest context the main concern is the need for modifying business mod-els. As only rarely a single model can serve different purposes, one expects that peoplewant to tailor a model to their own organisational context or even build an entirely newmodel. People having the knowledge to modify or build those models, usually don’thave the skills needed to implement it in a simulation engine. Thus what is neededis a fairly simple and easy to learn way of creating running simulations. The techni-cal skills we expect from users are more or less similar to the skills needed to usespreadsheets. Satisfying this requirement makes it possible that a business model canbe created or modified, and simulated interactively without any technical training. Somefurther, more detailed, requirements are:

– Dedicated support for creating, modifying and maintaining BMs, interventions andevents

– Vocabulary of the BM modeller and the tools is identical– No limits in terms of complexity of the model and all common mathematical mod-

elling constructs

– Automatic error detection where possible– Extensive support for simulation, visualisation and validation

We are not aware of an existing simulation environment that is sufficiently close towhat our model requires, in particular the notion of interventions and events acting onthe BM variables. Using a general purpose simulation or programming environment isnot an option, given the intended users.

From a historical point of view, we would like to note that in the KITS projectinitially extensive tool support was not deemed necessary. It soon became apparentthat the nature of BMs in general and the additional complexities of connecting KMinterventions to such BMs made dedicated tool support a necessity.

3 Architecture

In this section we provide an overview of the architecture of the tools. The main archi-tectural decision is to maintain two representations of the BM. The first representation isaspecificationin terms of the meta-model. This specification can be edited and browsedby a set of model entry tools. The second representation isoperational, and is a transla-tion of the specification into storage, computational statements and control structures.The operational representation, called the BM engine, computes successive states of theBM and is used by the simulation and validation tools. Obviously, the translation fromspecification to engine is completely transparent to the user (Fig. 2).

BMspecification

compilation

BMengine

model entrytools

simulationtools

Fig. 2.Representations and tools involved in the architecture of KMSIM

The requirements are realised in detail through seven tools which are briefly de-scribed below. There is an additional tool to make the BM engine available as a serverover the internet, following from a requirement of KM Quest. All tools are implementedin XPCE-Prolog [10].

Model entry (three tools) A BM can be created, modified and viewed using threemodel entry tools for the variables, interventions and events respectively (Sect. 4).

These tools allow the specification of the BM in terms the modeller will be fa-miliar with: status of variables, ranges, constraints on variable values, notions ofdecay and depreciation, influence over time, delay and effects of interventions andevents. In addition, some administrative aspects can be entered (domain within thecompany, precision for visualisation, description).

Charts design Charts are an important way to convey values of BM variables to theuser. A high-level chart design tool supports the definition of visually attractivecharts. The simulation tool automatically links values to the charts. The charts de-sign tool is not further discussed in this paper.

Simulation Interaction between a user (model developer, validator, game player) andthe BM is possible in a simulation tool (Sect. 5). The user can activate and de-activate interventions, issue events and view the effects on the BM variables ascharts, numerically (HTML, XML) or as a comprehensive visualisation of all knowl-edge process related variables (called theknowledge map).

Validation and tuning support A very important aspect of a simulation environmentis to provide assistance for tuning and validating the model. These tasks are sup-ported by a tuning tool (which randomly generates events and interventions andchecks whether user defined assertions are met) and a tool that traces the behaviourof the model graphically (Sect. 5.2).

Embedding A special version of the BM engine, called the BM server, can be run asa server on the internet in which it communicates using XML as input (specifyingevents and interventions) and outputs the BM state in XML and charts as bitmappedimages. The BM server is used as part of KM Quest.

BM interventions

events

BM entry and simulation tools

modeldeveloper

events and interventions

visualisationdefinitions

model state

visualisations

visualisationdesigner

validator /end user

model entry toolintervention toolevent tool

chart design tool

simulation tooltuning tool

validation tool

Fig. 3.Tools in KMSIM as seen from the roles of those interacting with them

Fig. 3 shows the roles of the various users involved with the tools. Themodel de-veloperuses the model entry tools to create a BM and associate KM interventions andevents with the BM. Avisualisation designerdefines how variables in the model areshown to the user. Thevalidator uses the the simulation, tuning and validation tools toverify the correctness of the BM. Often the model developer and the validator will be

the same person. And finally, theend userinteracts with the BM engine embedded inthe KM Quest game.

4 Model Entry Tools

The “core” BM represents a model of a company. The variables part of the BM arerelated to the business process (e.g. production level, number of employees), the knowl-edge process (e.g. competence in marketing), and the organisational effectiveness (profit,market share). The relations between these variables are such that the organisational ef-fectiveness of the company deteriorates (decay) when no attempts are made to improvethe knowledge process through interventions.

BP

I

KP

BP

I I

KPKP

BP

Intervention

Business process

Knowledge process

Input variables

Fig. 4. Interventions (and events) influence input variables, which in turn influence the knowledgeprocess variables and business process.

The link between the “core” BM and the knowledge management interventions andevents is represented by a set ofinput variables. A “complete” BM therefore consists ofthe “core” BM and the input variables (see also Fig. 4). Because there are no randomisedelements in the BM, it will always display the same behaviour when no interventionsare implemented and no events occur.

Interventions and events are defined in terms of how they affect the input variables.3

This makes it possible to define interventions and events independently of the BM. ABM that can be simulated therefore requires: (1) a BM consisting of variables repre-senting the business and knowledge processes, input variables that distribute the effects

3 External events can also affect the organisational effectiveness variables directly, for examplewhen a competitor brings an innovative product to the market and thereby gains market share.

of interventions and events over knowledge process variables; (2) a set of interventions;and (3) a set of events. Different simulations can be created by replacing the interven-tions and/or events, without changing the BM.

4.1 Business Model Entry Tool

The BM entry tool supports the creation and modification of a business model. Eachvariable in the model has several attributes (see Fig. 5) defined in an ontology. Most ofthe concepts in this ontology are fixed, some, for example the domains in the company(marketing, research) can be changed by the user. One of the attributes is aformulawhich explicitly specifies how the value of the variable is computed and implicitlydefines the influence relation between variables. An important aspect of the nature ofthe BM is the notion of time. This is modelled in discrete periods, which are calledcyclesin the tool.

The initial state of the model (cycle 0) is bootstrapped by computing constants. Theconstants represent a particular business case and can be used to “scale” the model forcompanies of various sizes or currencies.

Fig. 5.Business model entry tool

A subsequent state of the model is computed by ordering the variables on theirdependency on other variables. Once all dependent variables are computed, the formulaassociated with a variable is applied by the BM engine. Some examples of how BMnotions are mapped onto formulae are given below:

Decay As explained in Sect. 2 the decay of knowledge process variables is fundamen-tal to our meta-model. We can model the decay of knowledge utilisation with the

formula KU = KU * C1 (whereC1 is a constant, e.g.C1 = 0.94 ). The BMengine translates this formula toKUc = KUc−1 ∗ C1, whereKUc stands for thevalue ofKU in the current cycle andKUc−1 for the value in the previous cycle.Because of the propagation of values from knowledge process (via knowledge andbusiness process variables) to organisational effectiveness, the overall performanceof the company will also exhibit decay.

Propagation of influence An example of influence between variables is the formulafor competence in marketing:CM = KG + KD + KR. The level of competencedepends on knowledge gain (KG), development (KD) and retention (KR). Here, allvariables are computed using the same cycle, which implies that KG, KD and KRhave to be computed first and that the computation isCMc = KGc+KDc+KRc.A visualisation of influences and computation order can be seen in Fig. 8.

Relative change and delayed influenceRelative change and delayed influence can becomputed by referring to a previous cycle using the notationV - V[-1] whichis the difference between the value during the current and the previous cycle (com-putationallyVc − Vc−1).

Constrained values Values can be constrained by other values. For example, the saleslevel is constrained by the production level and sales level based on market share(see example in Fig. 5).

Depreciation Depreciation is an extreme case as it specifies decay in thefuture. For ex-ample, patents expire after some time. In the BM, depreciation is specified througha special function and the BM engine automatically subtracts the values for futurecycles.

Scaling and natural constraints The knowledge process variables are scaled to lie be-tween 1 and 10, this is specified in themin andmaxattributes of a variable. Simi-larly, market share has the natural constraint to lie between 0 and 100%.

Formulae may contain all common mathematical, conditional, comparison and log-ical operators.

The formulae are “hypotheses” about the relative dependencies that have to be testedby simulation, they are partly based on ideas from the literature. Defining the formulaesis obviously part of the knowledge acquisition problem the tools support.

4.2 Intervention and Event Entry Tools

Interventions (Fig. 6) and events consist of a control and a computational part. The con-trol part states whether the intervention or event is possible and the computational partstates which input variables (see Fig. 4) are affected. Events are slightly more compli-cated than interventions as they may depend on the current state of the model, whereasinterventions do not. For example, the event “intranet breaks down” requires that themodel is in a state in which the intranet is installed (through an intervention). Eventstherefore may have enabling and disabling conditions. If these conditions are not spec-ified, the event can always occur. Otherwise interventions and events are specified inprecisely the same manner and we will only consider interventions in this section.

The control and computational aspects of interventions are:

Fig. 6. Intervention entry tool

Control aspects The control aspects deal with the possibility, frequency and durationof an intervention. An intervention can be unavailable because it has already beenimplemented (e.g. installing an intranet). An intervention can be implemented alimited number of times (Max), there has to be some time between subsequent im-plementations (Periods) or the intervention is automatically removed after a certainnumber of cycles (Remove after).

Computational aspectsFor each input variable (middle browser in Fig. 6) affectedby an intervention, the effect has to be specified. Unfortunately, a simple formuladoes not suffice here, as the effect may be distributed over time in complicatedways. For example, installing an intranet has immediate effect on expenses (i.e.buying equipment), but there are additional expenses periodically (i.e. hiring staffto maintain the intranet). The effect of interventions (and events) is specified usingthe following vocabulary:Delay Many of the KM interventions do not take immediate effect, there is adelay

of some cycles.Initial effect The initial effect of an intervention is usually positive (e.g. knowl-

edge development increases).Next effect But, this effect disappears completely or partially.Repeat and repeat effectEffects can repeat every so many cycles (e.g. paying for

subscriptions).

The BM engine treats the effects of interventions similarly to depreciation. Whenan intervention is implemented the future changes to the input variables are computedand these values are used when computing subsequent states of the model. These futureeffects are not applied when an intervention is undone.

It may now also be apparent why existing simulation environments could not beused. The value of a variable depends on its formula, the cumulative effects of interven-tions and events, and depreciations.

5 Simulation and Validation

In the previous section we have described how to create a BM in KMSIM and how theBM engine computes subsequent states. Crunching out the numbers is, however, onlypart of the simulation.

Fig. 7. Simulation tool. The two clickable browsers on the left display the events and interven-tions. Colour coding is used to indicate whether they are possible or active. The results are shownon the right. At the bottom are controls to run a simulation.

5.1 Simulation

There are often several reasons for simulating a BM. The BM modeller needs simulationto study the effects of interventions and events on the behaviour of the model. The mainmotivation for simulation for the BM modeller is to validate and tune the model.

In KM Quest the learner is only provided with a partial picture. The learner canask for the past values ofoutput variables(organisational effectiveness and businessprocess) which are displayed using charts defined by the chart design tool. A completesimulation involves a little more. The state of the model at any point in time includesthe values of the BM variables and the status of interventions and events.

Fig. 7 shows the interactive simulation tool. The status of interventions and events(active, possible) is displayed using a colour coding scheme. When validating the modeldeveloper uses the simulation tool to implement interventions and to issue events. Thetool can visualise the BM variables in various ways: they can be shown in charts, asHTML tables (for later reference) or as a so calledknowledge map.

5.2 Validation and Tuning

Fig. 8. Influence graph. Vertices represent variables (I=input, S=state, O=output) and edges rep-resent influence. Colour is used to indicate whether the variable reflects organisational effective-ness, business process, knowledge process or knowledge.

Visualising the BM itself is mainly useful for the model developer. The most ob-vious visualisation is a graph that displays the influence relations between variables(Fig. 8). For compactness, the graph is displayed as a sphere where influence extendsinwards. Algorithms to draw such graphs dynamically can be found in [3]. The verticesin the graph are colour coded and indicate the status of the variable. The visualisa-tion makes the organisation of the meta-model clear. The outer ring contains the inputvariables (which are influenced by interventions and events), the second ring mainlycontains knowledge variables and the inner rings contain the business process and or-ganisational effectiveness variables (for some reasonProfit is in the centre). The graphhas turned out to be a very powerful tool for finding “obvious” errors in the model.

Fig. 9 shows a simulation of an intervention; note the use of KMSIM’s visualisationfeatures. The vertices again represent variables in the BM. On the left are the inputvariables affected by the selected intervention and all vertices are decorated with a

symbol indicating the value relative tonot implementing the intervention. For example,Profit (on the very right) is lower (H) as a result of the intervention, whereas most othervariables are higher (N) or the same (•).

Although the validation tools are aimed at validating and tuning the BM, they alsoturned out to be of practical value for finding errors in the implementation of the BMengine.

Fig. 9.Validating an intervention

6 Conclusions

The designer of tools is caught between the devil and the deep blue sea when faced withthe choice between generality and specificity. Making a tool very general increases itsapplicability but decreases its support for the user because it will contain less “content”about the application domain. Making it very specific decreases its applicability becauseit can only be used in a well defined limited context but increases its support for the userbecause it will contain more “content” about the application domain.

The natural tendency is to go for generality: this will appeal to a larger market. Asa consequence many simulation support tools are not too far removed from ordinary“visual” programming tools (e.g. PowersimR©4), which makes them still difficult to usefor domain experts without any programming experience. From a knowledge acquisi-tion and knowledge creation perspective these domain experts are the people who reallymatter. In domains where acquiring and creating knowledge by means of systematicempirical investigations is either very time consuming, hard or dangerous, simulation

4 It should be mentioned that this environment also has a kind of “meta-model”: system dynam-ics.

is the preferred way to validate theoretical models on their plausibility. So, supportingdomain experts in areas where these kind of limitations apply with easy-to-use tools forbuilding, inspecting and running simulated versions of their models, can be seen as akey area for knowledge acquisition. By necessity tools that serve this purpose will beon the “specific” side of the continuum outlined above. In our domain this specificity isderived from the nature of the domain, the nature of the models that must be build andthe intended users.

The domain described in this paper, knowledge management, and the tools devel-oped are a clear demonstration of the power of this approach. Of course, more experi-ence with the toolset is needed. For example, the range of users should be expanded,more research has to be done concerning actual ease of use, flexibility over a wide rangeof different business model types will be investigated. However, the application of thetool in the KITS project has significantly speeded up the creation and validation of acritical aspect of the learning environment: the business model represents in an activeway the company the learning is dealing with. At the same time, the availability of thetools will make creating versions of the business model, fitting very specific require-ments, much easier and this will contribute to the commercial value of the KM Questenvironment.

References

1. KMQuest simulation. See for more informationwww.kmquest.com .2. Tango! simulation. See for more informationwww.tangonow.net .3. G. Di Battista, P. Eades, R. Tamassia, and I. Tollis.Graph drawing: Algorithms for the

visualization of graphs. Prentice-Hall, Upper Sadle River, 1999.4. T.J. Beckman. The current state of knowledge management. In J. Liebowitz, editor,The

knowledge management handbook, pages 1.1–1.22. CRC Press, Boca Raton, 1999.5. T.H. Davenport. Knowledge management and the broader firm: strategy, advantage and

performance. In J. Liebowitz, editor,The knowledge management handbook, pages 2.1–2.11. CRC Press, Boca Raton, 1999.

6. A. Law and W. David Kelton.Simulation modeling and analysis. McGraw-Hill, Boston,2000.

7. A. Pritsker.Introduction to Simulation and SLAM II. John Wiley, New York, 1995.8. R. Reinhardt. Knowledge management: linking theory with practice. In D. Morey, M. May-

bury, and B. Thuraisingham, editors,Knowledge management: Classic and contemporarywork, pages 187–221. MIT Press, Cambridge, Mass, 2000.

9. A. Tiwana.The knowledge management toolkit: Practical techniques for building a knowl-edge management system. Prentice-Hall, Upper Sadle River, 2000.

10. J. Wielemaker and A. Anjewierden. Programming in XPCE-Prolog. Available fromwww.swi-prolog.org , 2002.